ARMWARE RFC Archive <- RFC Index (4401..4500)

RFC 4428


Network Working Group                              D. Papadimitriou, Ed.
Request for Comments: 4428                                       Alcatel
Category: Informational                                   E. Mannie, Ed.
                                                                Perceval
                                                              March 2006

 Analysis of Generalized Multi-Protocol Label Switching (GMPLS)-based
      Recovery Mechanisms (including Protection and Restoration)

Status of This Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2006).

Abstract

   This document provides an analysis grid to evaluate, compare, and
   contrast the Generalized Multi-Protocol Label Switching (GMPLS)
   protocol suite capabilities with the recovery mechanisms currently
   proposed at the IETF CCAMP Working Group.  A detailed analysis of
   each of the recovery phases is provided using the terminology defined
   in RFC 4427.  This document focuses on transport plane survivability
   and recovery issues and not on control plane resilience and related
   aspects.

Table of Contents

   1. Introduction ....................................................3
   2. Contributors ....................................................4
   3. Conventions Used in this Document ...............................5
   4. Fault Management ................................................5
      4.1. Failure Detection ..........................................5
      4.2. Failure Localization and Isolation .........................8
      4.3. Failure Notification .......................................9
      4.4. Failure Correlation .......................................11
   5. Recovery Mechanisms ............................................11
      5.1. Transport vs. Control Plane Responsibilities ..............11
      5.2. Technology-Independent and Technology-Dependent
           Mechanisms ................................................12
           5.2.1. OTN Recovery .......................................12
           5.2.2. Pre-OTN Recovery ...................................13
           5.2.3. SONET/SDH Recovery .................................13

Papadimitriou & Mannie       Informational                      [Page 1]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

      5.3. Specific Aspects of Control Plane-Based Recovery
           Mechanisms ................................................14
           5.3.1. In-Band vs. Out-Of-Band Signaling ..................14
           5.3.2. Uni- vs. Bi-Directional Failures ...................15
           5.3.3. Partial vs. Full Span Recovery .....................17
           5.3.4. Difference between LSP, LSP Segment and
                  Span Recovery ......................................18
      5.4. Difference between Recovery Type and Scheme ...............19
      5.5. LSP Recovery Mechanisms ...................................21
           5.5.1. Classification .....................................21
           5.5.2. LSP Restoration ....................................23
           5.5.3. Pre-Planned LSP Restoration ........................24
           5.5.4. LSP Segment Restoration ............................25
   6. Reversion ......................................................26
      6.1. Wait-To-Restore (WTR) .....................................26
      6.2. Revertive Mode Operation ..................................26
      6.3. Orphans ...................................................27
   7. Hierarchies ....................................................27
      7.1. Horizontal Hierarchy (Partitioning) .......................28
      7.2. Vertical Hierarchy (Layers) ...............................28
           7.2.1. Recovery Granularity ...............................30
      7.3. Escalation Strategies .....................................30
      7.4. Disjointness ..............................................31
           7.4.1. SRLG Disjointness ..................................32
   8. Recovery Mechanisms Analysis ...................................33
      8.1. Fast Convergence (Detection/Correlation and
           Hold-off Time) ............................................34
      8.2. Efficiency (Recovery Switching Time) ......................34
      8.3. Robustness ................................................35
      8.4. Resource Optimization .....................................36
           8.4.1. Recovery Resource Sharing ..........................37
           8.4.2. Recovery Resource Sharing and SRLG Recovery ........39
           8.4.3. Recovery Resource Sharing, SRLG
                  Disjointness and Admission Control .................40
   9. Summary and Conclusions ........................................42
   10. Security Considerations .......................................43
   11. Acknowledgements ..............................................43
   12. References ....................................................44
      12.1. Normative References .....................................44
      12.2. Informative References ...................................44

Papadimitriou & Mannie       Informational                      [Page 2]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

1.  Introduction

   This document provides an analysis grid to evaluate, compare, and
   contrast the Generalized MPLS (GMPLS) protocol suite capabilities
   with the recovery mechanisms proposed at the IETF CCAMP Working
   Group.  The focus is on transport plane survivability and recovery
   issues and not on control-plane-resilience-related aspects.  Although
   the recovery mechanisms described in this document impose different
   requirements on GMPLS-based recovery protocols, the protocols'
   specifications will not be covered in this document.  Though the
   concepts discussed are technology independent, this document
   implicitly focuses on SONET [T1.105]/SDH [G.707], Optical Transport
   Networks (OTN) [G.709], and pre-OTN technologies, except when
   specific details need to be considered (for instance, in the case of
   failure detection).

   A detailed analysis is provided for each of the recovery phases as
   identified in [RFC4427].  These phases define the sequence of generic
   operations that need to be performed when a LSP/Span failure (or any
   other event generating such failures) occurs:

      - Phase 1: Failure Detection
      - Phase 2: Failure Localization (and Isolation)
      - Phase 3: Failure Notification
      - Phase 4: Recovery (Protection or Restoration)
      - Phase 5: Reversion (Normalization)

   Together, failure detection, localization, and notification phases
   are referred to as "fault management".  Within a recovery domain, the
   entities involved during the recovery operations are defined in
   [RFC4427]; these entities include ingress, egress, and intermediate
   nodes.  The term "recovery mechanism" is used to cover both
   protection and restoration mechanisms.  Specific terms such as
   "protection" and "restoration" are used only when differentiation is
   required.  Likewise the term "failure" is used to represent both
   signal failure and signal degradation.

   In addition, when analyzing the different hierarchical recovery
   mechanisms including disjointness-related issues, a clear distinction
   is made between partitioning (horizontal hierarchy) and layering
   (vertical hierarchy).  In order to assess the current GMPLS protocol
   capabilities and the potential need for further extensions, the
   dimensions for analyzing each of the recovery mechanisms detailed in
   this document are introduced.  This document concludes by detailing
   the applicability of the current GMPLS protocol building blocks for
   recovery purposes.

Papadimitriou & Mannie       Informational                      [Page 3]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

2.  Contributors

   This document is the result of the CCAMP Working Group Protection and
   Restoration design team joint effort.  Besides the editors, the
   following are the authors that contributed to the present memo:

   Deborah Brungard (AT&T)
   200 S. Laurel Ave.
   Middletown, NJ 07748, USA

   EMail: dbrungard@att.com

   Sudheer Dharanikota

   EMail: sudheer@ieee.org

   Jonathan P. Lang (Sonos)
   506 Chapala Street
   Santa Barbara, CA 93101, USA

   EMail: jplang@ieee.org

   Guangzhi Li (AT&T)
   180 Park Avenue,
   Florham Park, NJ 07932, USA

   EMail: gli@research.att.com

   Eric Mannie
   Perceval
   Rue Tenbosch, 9
   1000 Brussels
   Belgium

   Phone: +32-2-6409194
   EMail: eric.mannie@perceval.net

   Dimitri Papadimitriou (Alcatel)
   Francis Wellesplein, 1
   B-2018 Antwerpen, Belgium

   EMail: dimitri.papadimitriou@alcatel.be

Papadimitriou & Mannie       Informational                      [Page 4]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   Bala Rajagopalan
   Microsoft India Development Center
   Hyderabad, India

   EMail: balar@microsoft.com

   Yakov Rekhter (Juniper)
   1194 N. Mathilda Avenue
   Sunnyvale, CA 94089, USA

   EMail: yakov@juniper.net

3.  Conventions Used in this Document

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED",  "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [RFC2119].

   Any other recovery-related terminology used in this document conforms
   to that defined in [RFC4427].  The reader is also assumed to be
   familiar with the terminology developed in [RFC3945], [RFC3471],
   [RFC3473], [RFC4202], and [RFC4204].

4.  Fault Management

4.1.  Failure Detection

   Transport failure detection is the only phase that cannot be achieved
   by the control plane alone because the latter needs a hook to the
   transport plane in order to collect the related information.  It has
   to be emphasized that even if failure events themselves are detected
   by the transport plane, the latter, upon a failure condition, must
   trigger the control plane for subsequent actions through the use of
   GMPLS signaling capabilities (see [RFC3471] and [RFC3473]) or Link
   Management Protocol capabilities (see [RFC4204], Section 6).

   Therefore, by definition, transport failure detection is transport
   technology dependent (and so exceptionally, we keep here the
   "transport plane" terminology).  In transport fault management,
   distinction is made between a defect and a failure.  Here, the
   discussion addresses failure detection (persistent fault cause).  In
   the technology-dependent descriptions, a more precise specification
   will be provided.

   As an example, SONET/SDH (see [G.707], [G.783], and [G.806]) provides
   supervision capabilities covering:

Papadimitriou & Mannie       Informational                      [Page 5]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   - Continuity: SONET/SDH monitors the integrity of the continuity of a
     trail (i.e., section or path).  This operation is performed by
     monitoring the presence/absence of the signal.  Examples are Loss
     of Signal (LOS) detection for the physical layer, Unequipped (UNEQ)
     Signal detection for the path layer, Server Signal Fail Detection
     (e.g., AIS) at the client layer.

   - Connectivity: SONET/SDH monitors the integrity of the routing of
     the signal between end-points.  Connectivity monitoring is needed
     if the layer provides flexible connectivity, either automatically
     (e.g., cross-connects) or manually (e.g., fiber distribution
     frame).  An example is the Trail (i.e., section or path) Trace
     Identifier used at the different layers and the corresponding Trail
     Trace Identifier Mismatch detection.

   - Alignment: SONET/SDH checks that the client and server layer frame
     start can be correctly recovered from the detection of loss of
     alignment.  The specific processes depend on the signal/frame
     structure and may include: (multi-)frame alignment, pointer
     processing, and alignment of several independent frames to a common
     frame start in case of inverse multiplexing.  Loss of alignment is
     a generic term.  Examples are loss of frame, loss of multi-frame,
     or loss of pointer.

   - Payload type: SONET/SDH checks that compatible adaptation functions
     are used at the source and the destination.  Normally, this is done
     by adding a payload type identifier (referred to as the "signal
     label") at the source adaptation function and comparing it with the
     expected identifier at the destination.  For instance, the payload
     type identifier is compared with the corresponding mismatch
     detection.

   - Signal Quality: SONET/SDH monitors the performance of a signal.
     For instance, if the performance falls below a certain threshold, a
     defect -- excessive errors (EXC) or degraded signal (DEG) -- is
     detected.

   The most important point is that the supervision processes and the
   corresponding failure detection (used to initiate the recovery
   phase(s)) result in either:

   - Signal Degrade (SD): A signal indicating that the associated data
     has degraded in the sense that a degraded defect condition is
     active (for instance, a dDEG declared when the Bit Error Rate
     exceeds a preset threshold).  Or

Papadimitriou & Mannie       Informational                      [Page 6]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   - Signal Fail (SF): A signal indicating that the associated data has
     failed in the sense that a signal interrupting near-end defect
     condition is active (as opposed to the degraded defect).

   In Optical Transport Networks (OTN), equivalent supervision
   capabilities are provided at the optical/digital section layers
   (i.e., Optical Transmission Section (OTS), Optical Multiplex Section
   (OMS) and Optical channel Transport Unit (OTU)) and at the
   optical/digital path layers (i.e., Optical Channel (OCh) and Optical
   channel Data Unit (ODU)).  Interested readers are referred to the
   ITU-T Recommendations [G.798] and [G.709] for more details.

   The above are examples that illustrate cases where the failure
   detection and reporting entities (see [RFC4427]) are co-located.  The
   following example illustrates the scenario where the failure
   detecting and reporting entities (see [RFC4427]) are not co-located.

   In pre-OTN networks, a failure may be masked by intermediate O-E-O
   based Optical Line System (OLS), preventing a Photonic Cross-Connect
   (PXC) from detecting upstream failures.  In such cases, failure
   detection may be assisted by an out-of-band communication channel,
   and failure condition may be reported to the PXC control plane.  This
   can be provided by using [RFC4209] extensions that deliver IP
   message-based communication between the PXC and the OLS control
   plane.  Also, since PXCs are independent of the framing format,
   failure conditions can only be triggered either by detecting the
   absence of the optical signal or by measuring its quality.  These
   mechanisms are generally less reliable than electrical (digital)
   ones.  Both types of detection mechanisms are outside the scope of
   this document.  If the intermediate OLS supports electrical (digital)
   mechanisms, using the LMP communication channel, these failure
   conditions are reported to

   the PXC and subsequent recovery actions are performed as described in
   Section 5.  As such, from the control plane viewpoint, this mechanism
   turns the OLS-PXC-composed system into a single logical entity, thus
   having the same failure management mechanisms as any other O-E-O
   capable device.

   More generally, the following are typical failure conditions in
   SONET/SDH and pre-OTN networks:

   - Loss of Light (LOL)/Loss of Signal (LOS): Signal Failure (SF)
     condition where the optical signal is not detected any longer on
     the receiver of a given interface.

   - Signal Degrade (SD): detection of the signal degradation over
     a specific period of time.

Papadimitriou & Mannie       Informational                      [Page 7]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   - For SONET/SDH payloads, all of the above-mentioned supervision
     capabilities can be used, resulting in SD or SF conditions.

   In summary, the following cases apply when considering the
   communication between the detecting and reporting entities:

   - Co-located detecting and reporting entities: both the detecting and
     reporting entities are on the same node (e.g., SONET/SDH equipment,
     Opaque cross-connects, and, with some limitations, Transparent
     cross-connects, etc.)

   - Non-co-located detecting and reporting entities:

     o with in-band communication between entities: entities are
       physically separated, but the transport plane provides in-band
       communication between them (e.g., Server Signal Failures such as
       Alarm Indication Signal (AIS), etc.)

     o with out-of-band communication between entities: entities are
       physically separated, but an out-of-band communication channel is
       provided between them (e.g., using [RFCF4204]).

4.2.  Failure Localization and Isolation

   Failure localization provides information to the deciding entity
   about the location (and so the identity) of the transport plane
   entity that detects the LSP(s)/span(s) failure.  The deciding entity
   can then make an accurate decision to achieve finer grained recovery
   switching action(s).  Note that this information can also be included
   as part of the failure notification (see Section 4.3).

   In some cases, this accurate failure localization information may be
   less urgent to determine if it requires performing more time-
   consuming failure isolation (see also Section 4.4).  This is
   particularly the case when edge-to-edge LSP recovery is performed
   based on a simple failure notification (including the identification
   of the working LSPs under failure condition).  Note that "edge"
   refers to a sub-network end-node, for instance.  In this case, a more
   accurate localization and isolation can be performed after recovery
   of these LSPs.

   Failure localization should be triggered immediately after the fault
   detection phase.  This operation can be performed at the transport
   plane and/or (if the operation is unavailable via the transport
   plane) the control plane level where dedicated signaling messages can
   be used.  When performed at the control plane level, a protocol such
   as LMP (see [RFC4204], Section 6) can be used for failure
   localization purposes.

Papadimitriou & Mannie       Informational                      [Page 8]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

4.3.  Failure Notification

   Failure notification is used 1) to inform intermediate nodes that an
   LSP/span failure has occurred and has been detected and 2) to inform
   the deciding entities (which can correspond to any intermediate or
   end-point of the failed LSP/span) that the corresponding service is
   not available.  In general, these deciding entities will be the ones
   making the appropriate recovery decision.  When co-located with the
   recovering entity, these entities will also perform the corresponding
   recovery action(s).

   Failure notification can be provided either by the transport or by
   the control plane.  As an example, let us first briefly describe the
   failure notification mechanism defined at the SONET/SDH transport
   plane level (also referred to as maintenance signal supervision):

   - AIS (Alarm Indication Signal) occurs as a result of a failure
     condition such as Loss of Signal and is used to notify downstream
     nodes (of the appropriate layer processing) that a failure has
     occurred.  AIS performs two functions: 1) inform the intermediate
     nodes (with the appropriate layer monitoring capability) that a
     failure has been detected and 2) notify the connection end-point
     that the service is no longer available.

   For a distributed control plane supporting one (or more) failure
   notification mechanism(s), regardless of the mechanism's actual
   implementation, the same capabilities are needed with more (or less)
   information provided about the LSPs/spans under failure condition,
   their detailed statuses, etc.

   The most important difference between these mechanisms is related to
   the fact that transport plane notifications (as defined today) would
   directly initiate either a certain type of protection switching (such
   as those described in [RFC4427]) via the transport plane or
   restoration actions via the management plane.

   On the other hand, using a failure notification mechanism through the
   control plane would provide the possibility of triggering either a
   protection or a restoration action via the control plane.  This has
   the advantage that a control-plane-recovery-responsible entity does
   not necessarily have to be co-located with a transport
   maintenance/recovery domain.  A control plane recovery domain can be
   defined at entities not supporting a transport plane recovery.

   Moreover, as specified in [RFC3473], notification message exchanges
   through a GMPLS control plane may not follow the same path as the
   LSP/spans for which these messages carry the status.  In turn, this
   ensures a fast, reliable (through acknowledgement and the use of

Papadimitriou & Mannie       Informational                      [Page 9]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   either a dedicated control plane network or disjoint control
   channels), and efficient (through the aggregation of several LSP/span
   statuses within the same message) failure notification mechanism.

   The other important properties to be met by the failure notification
   mechanism are mainly the following:

   - Notification messages must provide enough information such that the
     most efficient subsequent recovery action will be taken at the
     recovering entities (in most of the recovery types and schemes this
     action is even deterministic).  Remember here that these entities
     can be either intermediate or end-points through which normal
     traffic flows.  Based on local policy, intermediate nodes may not
     use this information for subsequent recovery actions (see for
     instance the APS protocol phases as described in [RFC4427]).  In
     addition, since fast notification is a mechanism running in
     collaboration with the existing GMPLS signaling (see [RFC3473])
     that also allows intermediate nodes to stay informed about the
     status of the working LSP/spans under failure condition.

     The trade-off here arises when defining what information the
     LSP/span end-points (more precisely, the deciding entities) need in
     order for the recovering entity to take the best recovery action:
     If not enough information is provided, the decision cannot be
     optimal (note that in this eventuality, the important issue is to
     quantify the level of sub-optimality).  If too much information is
     provided, the control plane may be overloaded with unnecessary
     information and the aggregation/correlation of this notification
     information will be more complex and time-consuming to achieve.
     Note that a more detailed quantification of the amount of
     information to be exchanged and processed is strongly dependent on
     the failure notification protocol.

   - If the failure localization and isolation are not performed by one
     of the LSP/span end-points or some intermediate points, the points
     should receive enough information from the notification message in
     order to locate the failure.  Otherwise, they would need to (re-)
     initiate a failure localization and isolation action.

   - Avoiding so-called notification storms implies that 1) the failure
     detection output is correlated (i.e., alarm correlation) and
     aggregated at the node detecting the failure(s), 2) the failure
     notifications are directed to a restricted set of destinations (in
     general the end-points), and 3) failure notification suppression
     (i.e., alarm suppression) is provided in order to limit flooding in
     case of multiple and/or correlated failures detected at several
     locations in the network.

Papadimitriou & Mannie       Informational                     [Page 10]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   - Alarm correlation and aggregation (at the failure-detecting node)
     implies a consistent decision based on the conditions for which a
     trade-off between fast convergence (at detecting node) and fast
     notification (implying that correlation and aggregation occurs at
     receiving end-points) can be found.

4.4.  Failure Correlation

   A single failure event (such as a span failure) can cause multiple
   failure (such as individual LSP failures) conditions to be reported.
   These can be grouped (i.e., correlated) to reduce the number of
   failure conditions communicated on the reporting channel, for both
   in-band and out-of-band failure reporting.

   In such a scenario, it can be important to wait for a certain period
   of time, typically called failure correlation time, and gather all
   the failures to report them as a group of failures (or simply group
   failure).  For instance, this approach can be provided using LMP-WDM
   for pre-OTN networks (see [RFC4209]) or when using Signal
   Failure/Degrade Group in the SONET/SDH context.

   Note that a default average time interval during which failure
   correlation operation can be performed is difficult to provide since
   it is strongly dependent on the underlying network topology.
   Therefore, providing a per-node configurable failure correlation time
   can be advisable.  The detailed selection criteria for this time
   interval are outside of the scope of this document.

   When failure correlation is not provided, multiple failure
   notification messages may be sent out in response to a single failure
   (for instance, a fiber cut).  Each failure notification message
   contains a set of information on the failed working resources (for
   instance, the individual lambda LSP flowing through this fiber).
   This allows for a more prompt response, but can potentially overload
   the control plane due to a large amount of failure notifications.

5.  Recovery Mechanisms

5.1.  Transport vs. Control Plane Responsibilities

   When applicable, recovery resources are provisioned, for both
   protection and restoration, using GMPLS signaling capabilities.
   Thus, these are control plane-driven actions (topological and
   resource-constrained) that are always performed in this context.

   The following tables give an overview of the responsibilities taken
   by the control plane in case of LSP/span recovery:

Papadimitriou & Mannie       Informational                     [Page 11]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   1. LSP/span Protection

   - Phase 1: Failure Detection                  Transport plane
   - Phase 2: Failure Localization/Isolation     Transport/Control plane
   - Phase 3: Failure Notification               Transport/Control plane
   - Phase 4: Protection Switching               Transport/Control plane
   - Phase 5: Reversion (Normalization)          Transport/Control plane

   Note: in the context of LSP/span protection, control plane actions
   can be performed either for operational purposes and/or
   synchronization purposes (vertical synchronization between transport
   and control plane) and/or notification purposes (horizontal
   synchronization between end-nodes at control plane level).  This
   suggests the selection of the responsible plane (in particular for
   protection switching) during the provisioning phase of the
   protected/protection LSP.

   2. LSP/span Restoration

   - Phase 1: Failure Detection                  Transport plane
   - Phase 2: Failure Localization/Isolation     Transport/Control plane
   - Phase 3: Failure Notification               Control plane
   - Phase 4: Recovery Switching                 Control plane
   - Phase 5: Reversion (Normalization)          Control plane

   Therefore, this document primarily focuses on provisioning of LSP
   recovery resources, failure notification mechanisms, recovery
   switching, and reversion operations.  Moreover, some additional
   considerations can be dedicated to the mechanisms associated to the
   failure localization/isolation phase.

5.2.  Technology-Independent and Technology-Dependent Mechanisms

   The present recovery mechanisms analysis applies to any circuit-
   oriented data plane technology with discrete bandwidth increments
   (like SONET/SDH, G.709 OTN, etc.) being controlled by a GMPLS-based
   distributed control plane.

   The following sub-sections are not intended to favor one technology
   versus another.  They list pro and cons for each technology in order
   to determine the mechanisms that GMPLS-based recovery must deliver to
   overcome their cons and make use of their pros in their respective
   applicability context.

5.2.1.  OTN Recovery

   OTN recovery specifics are left for further consideration.

Papadimitriou & Mannie       Informational                     [Page 12]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

5.2.2.  Pre-OTN Recovery

   Pre-OTN recovery specifics (also referred to as "lambda switching")
   present mainly the following advantages:

   - They benefit from a simpler architecture, making it more suitable
     for mesh-based recovery types and schemes (on a per-channel basis).

   - Failure suppression at intermediate node transponders, e.g., use of
     squelching, implies that failures (such as LoL) will propagate to
     edge nodes.  Thus, edge nodes will have the possibility to initiate
     recovery actions driven by upper layers (vs. use of non-standard
     masking of upstream failures).

   The main disadvantage is the lack of interworking due to the large
   amount of failure management (in particular failure notification
   protocols) and recovery mechanisms currently available.

   Note also, that for all-optical networks, combination of recovery
   with optical physical impairments is left for a future release of
   this document because corresponding detection technologies are under
   specification.

5.2.3.  SONET/SDH Recovery

   Some of the advantages of SONET [T1.105]/SDH [G.707], and more
   generically any Time Division Multiplexing (TDM) transport plane
   recovery, are that they provide:

   - Protection types operating at the data plane level that are
     standardized (see [G.841]) and can operate across protected domains
     and interwork (see [G.842]).

   - Failure detection, notification, and path/section Automatic
     Protection Switching (APS) mechanisms.

   - Greater control over the granularity of the TDM LSPs/links that can
     be recovered with respect to coarser optical channel (or whole
     fiber content) recovery switching

   Some of the limitations of the SONET/SDH recovery are:

   - Limited topological scope: Inherently the use of ring topologies,
     typically, dedicated Sub-Network Connection Protection (SNCP) or
     shared protection rings, has reduced flexibility and resource
     efficiency with respect to the (somewhat more complex) meshed
     recovery.

Papadimitriou & Mannie       Informational                     [Page 13]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   - Inefficient use of spare capacity: SONET/SDH protection is largely
     applied to ring topologies, where spare capacity often remains
     idle, making the efficiency of bandwidth usage a real issue.

   - Support of meshed recovery requires intensive network management
     development, and the functionality is limited by both the network
     elements and the capabilities of the element management systems
     (thus justifying the development of GMPLS-based distributed
     recovery mechanisms.)

5.3.  Specific Aspects of Control Plane-Based Recovery Mechanisms

5.3.1.  In-Band vs. Out-Of-Band Signaling

   The nodes communicate through the use of IP terminating control
   channels defining the control plane (transport) topology.  In this
   context, two classes of transport mechanisms can be considered here:
   in-fiber or out-of-fiber (through a dedicated physically diverse
   control network referred to as the Data Communication Network or
   DCN).  The potential impact of the usage of an in-fiber (signaling)
   transport mechanism is briefly considered here.

   In-fiber transport mechanisms can be further subdivided into in-band
   and out-of-band.  As such, the distinction between in-fiber in-band
   and in-fiber out-of-band signaling reduces to the consideration of a
   logically- versus physically-embedded control plane topology with
   respect to the transport plane topology.  In the scope of this
   document, it is assumed that at least one IP control channel between
   each pair of adjacent nodes is continuously available to enable the
   exchange of recovery-related information and messages.  Thus, in
   either case (i.e., in-band or out-of-band) at least one logical or
   physical control channel between each pair of nodes is always
   expected to be available.

   Therefore, the key issue when using in-fiber signaling is whether one
   can assume independence between the fault-tolerance capabilities of
   control plane and the failures affecting the transport plane
   (including the nodes).  Note also that existing specifications like
   the OTN provide a limited form of independence for in-fiber signaling
   by dedicating a separate optical supervisory channel (OSC, see
   [G.709] and [G.874]) to transport the overhead and other control
   traffic.  For OTNs, failure of the OSC does not result in failing the
   optical channels.  Similarly, loss of the control channel must not
   result in failing the data channels (transport plane).

Papadimitriou & Mannie       Informational                     [Page 14]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

5.3.2.  Uni- vs. Bi-Directional Failures

   The failure detection, correlation, and notification mechanisms
   (described in Section 4) can be triggered when either a uni-
   directional or a bi-directional LSP/Span failure occurs (or a
   combination of both).  As illustrated in Figures 1 and 2, two
   alternatives can be considered here:

   1. Uni-directional failure detection: the failure is detected on the
      receiver side, i.e., it is detected by only the downstream node to
      the failure (or by the upstream node depending on the failure
      propagation direction, respectively).

   2. Bi-directional failure detection: the failure is detected on the
      receiver side of both downstream node AND upstream node to the
      failure.

   Notice that after the failure detection time, if only control-plane-
   based failure management is provided, the peering node is unaware of
   the failure detection status of its neighbor.

    -------             -------           -------             -------
   |       |           |       |Tx     Rx|       |           |       |
   | NodeA |----...----| NodeB |xxxxxxxxx| NodeC |----...----| NodeD |
   |       |----...----|       |---------|       |----...----|       |
    -------             -------           -------             -------

   t0                                >>>>>>> F

   t1                      x <---------------x
                               Notification
   t2  <--------...--------x                 x--------...-------->
          Up Notification                      Down Notification

              Figure 1: Uni-directional failure detection

Papadimitriou & Mannie       Informational                     [Page 15]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

    -------             -------           -------             -------
   |       |           |       |Tx     Rx|       |           |       |
   | NodeA |----...----| NodeB |xxxxxxxxx| NodeC |----...----| NodeD |
   |       |----...----|       |xxxxxxxxx|       |----...----|       |
    -------             -------           -------             -------

   t0                      F <<<<<<< >>>>>>> F

   t1                      x <-------------> x
                               Notification
   t2  <--------...--------x                 x--------...-------->
          Up Notification                      Down Notification

               Figure 2: Bi-directional failure detection

   After failure detection, the following failure management operations
   can be subsequently considered:

   - Each detecting entity sends a notification message to the
     corresponding transmitting entity.  For instance, in Figure 1, node
     C sends a notification message to node B.  In Figure 2, node C
     sends a notification message to node B while node B sends a
     notification message to node C.  To ensure reliable failure
     notification, a dedicated acknowledgement message can be returned
     back to the sender node.

   - Next, within a certain (and pre-determined) time window, nodes
     impacted by the failure occurrences may perform their correlation.
     In case of uni-directional failure, node B only receives the
     notification message from node C, and thus the time for this
     operation is negligible.  In case of bi-directional failure, node B
     has to correlate the received notification message from node C with
     the corresponding locally detected information (and node C has to
     do the same with the message from node B).

   - After some (pre-determined) period of time, referred to as the
     hold-off time, if the local recovery actions (see Section 5.3.4)
     were not successful, the following occurs.  In case of uni-
     directional failure and depending on the directionality of the LSP,
     node B should send an upstream notification message (see [RFC3473])
     to the ingress node A.  Node C may send a downstream notification
     message (see [RFC3473]) to the egress node D.  However, in that
     case, only node A would initiate an edge to edge recovery action.
     Node A is referred to as the "master", and node D is referred to as
     the "slave", per [RFC4427].  Note that the other LSP end-node (node
     D in this case) may be optionally notified using a downstream
     notification message (see [RFC3473]).

Papadimitriou & Mannie       Informational                     [Page 16]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

     In case of bi-directional failure, node B should send an upstream
     notification message (see [RFC3473]) to the ingress node A.  Node C
     may send a downstream notification message (see [RFC3473]) to the
     egress node D.  However, due to the dependence on the LSP
     directionality, only ingress node A would initiate an edge-to-edge
     recovery action.  Note that the other LSP end-node (node D in this
     case) should also be notified of this event using a downstream
     notification message (see [RFC3473]).  For instance, if an LSP
     directed from D to A is under failure condition, only the
     notification message sent from node C to D would initiate a
     recovery action.  In this case, per [RFC4427], the deciding and
     recovering node D is referred to as the "master", while node A is
     referred to as the "slave" (i.e., recovering only entity).

     Note: The determination of the master and the slave may be based
     either on configured information or dedicated protocol capability.

   In the above scenarios, the path followed by the upstream and
   downstream notification messages does not have to be the same as the
   one followed by the failed LSP (see [RFC3473] for more details on the
   notification message exchange).  The important point concerning this
   mechanism is that either the detecting/reporting entity (i.e., nodes
   B and C) is also the deciding/recovery entity or the
   detecting/reporting entity is simply an intermediate node in the
   subsequent recovery process.  One refers to local recovery in the
   former case, and to edge-to-edge recovery in the latter one (see also
   Section 5.3.4).

5.3.3.  Partial vs. Full Span Recovery

   When a given span carries more than one LSP or LSP segment, an
   additional aspect must be considered.  In case of span failure, the
   LSPs it carries can be recovered individually, as a group (aka bulk
   LSP recovery), or as independent sub-groups.  When correlation time
   windows are used and simultaneous recovery of several LSPs can be
   performed using a single request, the selection of this mechanism
   would be triggered independently of the failure notification
   granularity.  Moreover, criteria for forming such sub-groups are
   outside of the scope of this document.

   Additional complexity arises in the case of (sub-)group LSP recovery.
   Between a given pair of nodes, the LSPs that a given (sub-)group
   contains may have been created from different source nodes (i.e.,
   initiator) and directed toward different destination nodes.
   Consequently the failure notification messages following a bi-
   directional span failure that affects several LSPs (or the whole
   group of LSPs it carries) are not necessarily directed toward the
   same initiator nodes.  In particular, these messages may be directed

Papadimitriou & Mannie       Informational                     [Page 17]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   to both the upstream and downstream nodes to the failure.  Therefore,
   such span failure may trigger recovery actions to be performed from
   both sides (i.e., from both the upstream and the downstream nodes to
   the failure).  In order to facilitate the definition of the
   corresponding recovery mechanisms (and their sequence), one assumes
   here as well that, per [RFC4427], the deciding (and recovering)
   entity (referred to as the "master") is the only initiator of the
   recovery of the whole LSP (sub-)group.

5.3.4.  Difference between LSP, LSP Segment and Span Recovery

   The recovery definitions given in [RFC4427] are quite generic and
   apply for link (or local span) and LSP recovery.  The major
   difference between LSP, LSP Segment and span recovery is related to
   the number of intermediate nodes that the signaling messages have to
   travel.  Since nodes are not necessarily adjacent in the case of LSP
   (or LSP Segment) recovery, signaling message exchanges from the
   reporting to the deciding/recovery entity may have to cross several
   intermediate nodes.  In particular, this applies to the notification
   messages due to the number of hops separating the location of a
   failure occurrence from its destination.  This results in an
   additional propagation and forwarding delay.  Note that the former
   delay may in certain circumstances be non-negligible; e.g., in a
   copper out-of-band network, the delay is approximately 1 ms per
   200km.

   Moreover, the recovery mechanisms applicable to end-to-end LSPs and
   to the segments that may compose an end-to-end LSP (i.e., edge-to-
   edge recovery) can be exactly the same.  However, one expects in the
   latter case, that the destination of the failure notification message
   will be the ingress/egress of each of these segments.  Therefore,
   using the mechanisms described in Section 5.3.2, failure notification
   messages can be exchanged first between terminating points of the LSP
   segment, and after expiration of the hold-off time, between
   terminating points of the end-to-end LSP.

   Note: Several studies provide quantitative analysis of the relative
   performance of LSP/span recovery techniques. [WANG] for instance,
   provides an analysis grid for these techniques showing that dynamic
   LSP restoration (see Section 5.5.2) performs well under medium
   network loads, but suffers performance degradations at higher loads
   due to greater contention for recovery resources.  LSP restoration
   upon span failure, as defined in [WANG], degrades at higher loads
   because paths around failed links tend to increase the hop count of
   the affected LSPs and thus consume additional network resources.
   Also, performance of LSP restoration can be enhanced by a failed
   working LSP's source node that initiates a new recovery attempt if an
   initial attempt fails.  A single retry attempt is sufficient to

Papadimitriou & Mannie       Informational                     [Page 18]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   produce large increases in the restoration success rate and ability
   to initiate successful LSP restoration attempts, especially at high
   loads, while not adding significantly to the long-term average
   recovery time.  Allowing additional attempts produces only small
   additional gains in performance.  This suggests using additional
   (intermediate) crankback signaling when using dynamic LSP restoration
   (described in Section 5.5.2 - case 2).  Details on crankback
   signaling are outside the scope of this document.

5.4.  Difference between Recovery Type and Scheme

   [RFC4427] defines the basic LSP/span recovery types.  This section
   describes the recovery schemes that can be built using these recovery
   types.  In brief, a recovery scheme is defined as the combination of
   several ingress-egress node pairs supporting a given recovery type
   (from the set of the recovery types they allow).  Several examples
   are provided here to illustrate the difference between recovery types
   such as 1:1 or M:N, and recovery schemes such as (1:1)^n or (M:N)^n
   (referred to as shared-mesh recovery).

   1. (1:1)^n with recovery resource sharing

   The exponent, n, indicates the number of times a 1:1 recovery type is
   applied between at most n different ingress-egress node pairs.  Here,
   at most n pairs of disjoint working and recovery LSPs/spans share a
   common resource at most n times.  Since the working LSPs/spans are
   mutually disjoint, simultaneous requests for use of the shared
   (common) resource will only occur in case of simultaneous failures,
   which are less likely to happen.

   For instance, in the common (1:1)^2 case, if the 2 recovery LSPs in
   the group overlap the same common resource, then it can handle only
   single failures; any multiple working LSP failures will cause at
   least one working LSP to be denied automatic recovery.  Consider for
   instance the following topology with the working LSPs A-B-C and F-G-H
   and their respective recovery LSPs A-D-E-C and F-D-E-H that share a
   common D-E link resource.

                          A---------B---------C
                           \                 /
                            \               /
                             D-------------E
                            /               \
                           /                 \
                          F---------G---------H

Papadimitriou & Mannie       Informational                     [Page 19]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   2. (M:N)^n with recovery resource sharing

   The (M:N)^n scheme is documented here for the sake of completeness
   only (i.e., it is not mandated that GMPLS capabilities support this
   scheme).  The exponent, n, indicates the number of times an M:N
   recovery type is applied between at most n different ingress-egress
   node pairs.  So the interpretation follows from the previous case,
   except that here disjointness applies to the N working LSPs/spans and
   to the M recovery LSPs/spans while sharing at most n times M common
   resources.

   In both schemes, it results in a "group" of sum{n=1}^N N{n} working
   LSPs and a pool of shared recovery resources, not all of which are
   available to any given working LSP.  In such conditions, defining a
   metric that describes the amount of overlap among the recovery LSPs
   would give some indication of the group's ability to handle
   simultaneous failures of multiple LSPs.

   For instance, in the simple (1:1)^n case, if n recovery LSPs in a
   (1:1)^n group overlap, then the group can handle only single
   failures; any simultaneous failure of multiple working LSPs will
   cause at least one working LSP to be denied automatic recovery.  But
   if one considers, for instance, a (2:2)^2 group in which there are
   two pairs of overlapping recovery LSPs, then two LSPs (belonging to
   the same pair) can be simultaneously recovered.  The latter case can
   be illustrated by the following topology with 2 pairs of working LSPs
   A-B-C and F-G-H and their respective recovery LSPs A-D-E-C and
   F-D-E-H that share two common D-E link resources.

                           A========B========C
                           \\               //
                            \\             //
                             D =========== E
                            //             \\
                           //               \\
                           F========G========H

   Moreover, in all these schemes, (working) path disjointness can be
   enforced by exchanging information related to working LSPs during the
   recovery LSP signaling.  Specific issues related to the combination
   of shared (discrete) bandwidth and disjointness for recovery schemes
   are described in Section 8.4.2.

Papadimitriou & Mannie       Informational                     [Page 20]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

5.5.  LSP Recovery Mechanisms

5.5.1.  Classification

   The recovery time and ratio of LSPs/spans depend on proper recovery
   LSP provisioning (meaning pre-provisioning when performed before
   failure occurrence) and the level of overbooking of recovery
   resources (i.e., over-provisioning).  A proper balance of these two
   operations will result in the desired LSP/span recovery time and
   ratio when single or multiple failures occur.  Note also that these
   operations are mostly performed during the network planning phases.

   The different options for LSP (pre-)provisioning and overbooking are
   classified below to structure the analysis of the different recovery
   mechanisms.

   1. Pre-Provisioning

   Proper recovery LSP pre-provisioning will help to alleviate the
   failure of the working LSPs (due to the failure of the resources that
   carry these LSPs).  As an example, one may compute and establish the
   recovery LSP either end-to-end or segment-per-segment, to protect a
   working LSP from multiple failure events affecting link(s), node(s)
   and/or SRLG(s).  The recovery LSP pre-provisioning options are
   classified as follows in the figure below:

   (1) The recovery path can be either pre-computed or computed on-
       demand.

   (2) When the recovery path is pre-computed, it can be either pre-
       signaled (implying recovery resource reservation) or signaled
       on-demand.

   (3) When the recovery resources are pre-signaled, they can be either
       pre-selected or selected on-demand.

   Recovery LSP provisioning phases:

Papadimitriou & Mannie       Informational                     [Page 21]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   (1) Path Computation --> On-demand
           |
           |
            --> Pre-Computed
                    |
                    |
                   (2) Signaling --> On-demand
                           |
                           |
                            --> Pre-Signaled
                                    |
                                    |
                                   (3) Resource Selection --> On-demand
                                                |
                                                |
                                                 --> Pre-Selected

   Note that these different options lead to different LSP/span recovery
   times.  The following sections will consider the above-mentioned
   pre-provisioning options when analyzing the different recovery
   mechanisms.

   2. Overbooking

   There are many mechanisms available that allow the overbooking of the
   recovery resources.  This overbooking can be done per LSP (as in the
   example mentioned above), per link (such as span protection), or even
   per domain.  In all these cases, the level of overbooking, as shown
   in the below figure, can be classified as dedicated (such as 1+1 and
   1:1), shared (such as 1:N and M:N), or unprotected (and thus
   restorable, if enough recovery resources are available).

   Overbooking levels:

                    +----- Dedicated (for instance: 1+1, 1:1, etc.)
                    |
                    |

                    +----- Shared (for instance: 1:N, M:N, etc.)
                    |
   Level of         |
   Overbooking -----+----- Unprotected (for instance: 0:1, 0:N)

   Also, when using shared recovery, one may support preemptible extra-
   traffic; the recovery mechanism is then expected to allow preemption
   of this low priority traffic in case of recovery resource contention
   during recovery operations.  The following sections will consider the

Papadimitriou & Mannie       Informational                     [Page 22]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   above-mentioned overbooking options when analyzing the different
   recovery mechanisms.

5.5.2.  LSP Restoration

   The following times are defined to provide a quantitative estimation
   about the time performance of the different LSP restoration
   mechanisms (also referred to as LSP re-routing):

   - Path Computation Time: Tc
   - Path Selection Time: Ts
   - End-to-end LSP Resource Reservation Time: Tr (a delta for resource
     selection is also considered, the corresponding total time is then
     referred to as Trs)
   - End-to-end LSP Resource Activation Time: Ta (a delta for
     resource selection is also considered, the corresponding total
     time is then referred to as Tas)

   The Path Selection Time (Ts) is considered when a pool of recovery
   LSP paths between a given pair of source/destination end-points is
   pre-computed, and after a failure occurrence one of these paths is
   selected for the recovery of the LSP under failure condition.

   Note: failure management operations such as failure detection,
   correlation, and notification are considered (for a given failure
   event) as equally time-consuming for all the mechanisms described
   below:

   1. With Route Pre-computation (or LSP re-provisioning)

   An end-to-end restoration LSP is established after the failure(s)
   occur(s) based on a pre-computed path.  As such, one can define this
   as an "LSP re-provisioning" mechanism.  Here, one or more (disjoint)
   paths for the restoration LSP are computed (and optionally pre-
   selected) before a failure occurs.

   No reservation or selection of resources is performed along the
   restoration path before failure occurrence.  As a result, there is no
   guarantee that a restoration LSP is available when a failure occurs.

   The expected total restoration time T is thus equal to Ts + Trs or to
   Trs when a dedicated computation is performed for each working LSP.

   2. Without Route Pre-computation (or Full LSP re-routing)

   An end-to-end restoration LSP is dynamically established after the
   failure(s) occur(s).  After failure occurrence, one or more
   (disjoint) paths for the restoration LSP are dynamically computed and

Papadimitriou & Mannie       Informational                     [Page 23]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   one is selected.  As such, one can define this as a complete "LSP
   re-routing" mechanism.

   No reservation or selection of resources is performed along the
   restoration path before failure occurrence.  As a result, there is no
   guarantee that a restoration LSP is available when a failure occurs.

   The expected total restoration time T is thus equal to Tc (+ Ts) +
   Trs.  Therefore, time performance between these two approaches
   differs by the time required for route computation Tc (and its
   potential selection time, Ts).

5.5.3.  Pre-Planned LSP Restoration

   Pre-planned LSP restoration (also referred to as pre-planned LSP re-
   routing) implies that the restoration LSP is pre-signaled.  This in
   turn implies the reservation of recovery resources along the
   restoration path.  Two cases can be defined based on whether the
   recovery resources are pre-selected.

   1. With resource reservation and without resource pre-selection

   Before failure occurrence, an end-to-end restoration path is pre-
   selected from a set of pre-computed (disjoint) paths.  The
   restoration LSP is signaled along this pre-selected path to reserve
   resources at each node, but these resources are not selected.

   In this case, the resources reserved for each restoration LSP may be
   dedicated or shared between multiple restoration LSPs whose working
   LSPs are not expected to fail simultaneously.  Local node policies
   can be applied to define the degree to which these resources can be
   shared across independent failures.  Also, since a restoration scheme
   is considered, resource sharing should not be limited to restoration
   LSPs that start and end at the same ingress and egress nodes.
   Therefore, each node participating in this scheme is expected to
   receive some feedback information on the sharing degree of the
   recovery resource(s) that this scheme involves.

   Upon failure detection/notification message reception, signaling is
   initiated along the restoration path to select the resources, and to
   perform the appropriate operation at each node crossed by the
   restoration LSP (e.g., cross-connections).  If lower priority LSPs
   were established using the restoration resources, they must be
   preempted when the restoration LSP is activated.

   Thus, the expected total restoration time T is equal to Tas (post-
   failure activation), while operations performed before failure
   occurrence take Tc + Ts + Tr.

Papadimitriou & Mannie       Informational                     [Page 24]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   2. With both resource reservation and resource pre-selection

   Before failure occurrence, an end-to-end restoration path is pre-
   selected from a set of pre-computed (disjoint) paths.  The
   restoration LSP is signaled along this pre-selected path to reserve
   AND select resources at each node, but these resources are not
   committed at the data plane level.  So that the selection of the
   recovery resources is committed at the control plane level only, no
   cross-connections are performed along the restoration path.

   In this case, the resources reserved and selected for each
   restoration LSP may be dedicated or even shared between multiple
   restoration LSPs whose associated working LSPs are not expected to
   fail simultaneously.  Local node policies can be applied to define
   the degree to which these resources can be shared across independent
   failures.  Also, because a restoration scheme is considered, resource
   sharing should not be limited to restoration LSPs that start and end
   at the same ingress and egress nodes.  Therefore, each node
   participating in this scheme is expected to receive some feedback
   information on the sharing degree of the recovery resource(s) that
   this scheme involves.

   Upon failure detection/notification message reception, signaling is
   initiated along the restoration path to activate the reserved and
   selected resources, and to perform the appropriate operation at each
   node crossed by the restoration LSP (e.g., cross-connections).  If
   lower priority LSPs were established using the restoration resources,
   they must be preempted when the restoration LSP is activated.

   Thus, the expected total restoration time T is equal to Ta (post-
   failure activation), while operations performed before failure
   occurrence take Tc + Ts + Trs.  Therefore, time performance between
   these two approaches differs only by the time required for resource
   selection during the activation of the recovery LSP (i.e., Tas - Ta).

5.5.4.  LSP Segment Restoration

   The above approaches can be applied on an edge-to-edge LSP basis
   rather than end-to-end LSP basis (i.e., to reduce the global recovery
   time) by allowing the recovery of the individual LSP segments
   constituting the end-to-end LSP.

   Also, by using the horizontal hierarchy approach described in Section
   7.1, an end-to-end LSP can be recovered by multiple recovery
   mechanisms applied on an LSP segment basis (e.g., 1:1 edge-to-edge
   LSP protection in a metro network, and M:N edge-to-edge protection in
   the core).  These mechanisms are ideally independent and may even use
   different failure localization and notification mechanisms.

Papadimitriou & Mannie       Informational                     [Page 25]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

6.  Reversion

   Reversion (a.k.a. normalization) is defined as the mechanism allowing
   switching of normal traffic from the recovery LSP/span to the working
   LSP/span previously under failure condition.  Use of normalization is
   at the discretion of the recovery domain policy.  Normalization may
   impact the normal traffic (a second hit) depending on the
   normalization mechanism used.

   If normalization is supported, then 1) the LSP/span must be returned
   to the working LSP/span when the failure condition clears and 2) the
   capability to de-activate (turn-off) the use of reversion should be
   provided.  De-activation of reversion should not impact the normal
   traffic, regardless of whether it is currently using the working or
   recovery LSP/span.

   Note: during the failure, the reuse of any non-failed resources
   (e.g., LSP and/or spans) belonging to the working LSP/span is under
   the discretion of recovery domain policy.

6.1.  Wait-To-Restore (WTR)

   A specific mechanism (Wait-To-Restore) is used to prevent frequent
   recovery switching operations due to an intermittent defect (e.g.,
   Bit Error Rate (BER) fluctuating around the SD threshold).

   First, an LSP/span under failure condition must become fault-free,
   e.g., a BER less than a certain recovery threshold.  After the
   recovered LSP/span (i.e., the previously working LSP/span) meets this
   criterion, a fixed period of time shall elapse before normal traffic
   uses the corresponding resources again.  This duration called Wait-
   To-Restore (WTR) period or timer is generally on the order of a few
   minutes (for instance, 5 minutes) and should be capable of being set.
   The WTR timer may be either a fixed period, or provide for
   incrementally longer periods before retrying.  An SF or SD condition
   on the previously working LSP/span will override the WTR timer value
   (i.e., the WTR cancels and the WTR timer will restart).

6.2.  Revertive Mode Operation

   In revertive mode of operation, when the recovery LSP/span is no
   longer required, i.e., the failed working LSP/span is no longer in SD
   or SF condition, a local Wait-to-Restore (WTR) state will be
   activated before switching the normal traffic back to the recovered
   working LSP/span.

   During the reversion operation, since this state becomes the highest
   in priority, signaling must maintain the normal traffic on the

Papadimitriou & Mannie       Informational                     [Page 26]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   recovery LSP/span from the previously failed working LSP/span.
   Moreover, during this WTR state, any null traffic or extra traffic
   (if applicable) request is rejected.

   However, deactivation (cancellation) of the wait-to-restore timer may
   occur if there are higher priority request attempts.  That is, the
   recovery LSP/span usage by the normal traffic may be preempted if a
   higher priority request for this recovery LSP/span is attempted.

6.3.  Orphans

   When a reversion operation is requested, normal traffic must be
   switched from the recovery to the recovered working LSP/span.  A
   particular situation occurs when the previously working LSP/span
   cannot be recovered, so normal traffic cannot be switched back.  In
   that case, the LSP/span under failure condition (also referred to as
   "orphan") must be cleared (i.e., removed) from the pool of resources
   allocated for normal traffic.  Otherwise, potential de-
   synchronization between the control and transport plane resource
   usage can appear.  Depending on the signaling protocol capabilities
   and behavior, different mechanisms are expected here.

   Therefore, any reserved or allocated resources for the LSP/span under
   failure condition must be unreserved/de-allocated.  Several ways can
   be used for that purpose: wait for the clear-out time interval to
   elapse, initiate a deletion from the ingress or the egress node, or
   trigger the initiation of deletion from an entity (such as an EMS or
   NMS) capable of reacting upon reception of an appropriate
   notification message.

7.  Hierarchies

   Recovery mechanisms are being made available at multiple (if not all)
   transport layers within so-called "IP/MPLS-over-optical" networks.
   However, each layer has certain recovery features, and one needs to
   determine the exact impact of the interaction between the recovery
   mechanisms provided by these layers.

   Hierarchies are used to build scalable complex systems.  By hiding
   the internal details, abstraction is used as a mechanism to build
   large networks or as a technique for enforcing technology,
   topological, or administrative boundaries.  The same hierarchical
   concept can be applied to control the network survivability.  Network
   survivability is the set of capabilities that allow a network to
   restore affected traffic in the event of a failure.  Network
   survivability is defined further in [RFC4427].  In general, it is
   expected that the recovery action is taken by the recoverable
   LSP/span closest to the failure in order to avoid the multiplication

Papadimitriou & Mannie       Informational                     [Page 27]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   of recovery actions.  Moreover, recovery hierarchies also can be
   bound to control plane logical partitions (e.g., administrative or
   topological boundaries).  Each logical partition may apply different
   recovery mechanisms.

   In brief, it is commonly accepted that the lower layers can provide
   coarse but faster recovery while the higher layers can provide finer
   but slower recovery.  Moreover, it is also desirable to avoid similar
   layers with functional overlaps in order to optimize network resource
   utilization and processing overhead, since repeating the same
   capabilities at each layer does not create any added value for the
   network as a whole.  In addition, even if a lower layer recovery
   mechanism is enabled, it does not prevent the additional provision of
   a recovery mechanism at the upper layer.  The inverse statement does
   not necessarily hold; that is, enabling an upper layer recovery
   mechanism may prevent the use of a lower layer recovery mechanism.
   In this context, this section analyzes these hierarchical aspects
   including the physical (passive) layer(s).

7.1.  Horizontal Hierarchy (Partitioning)

   A horizontal hierarchy is defined when partitioning a single-layer
   network (and its control plane) into several recovery domains.
   Within a domain, the recovery scope may extend over a link (or span),
   LSP segment, or even an end-to-end LSP.  Moreover, an administrative
   domain may consist of a single recovery domain or can be partitioned
   into several smaller recovery domains.  The operator can partition
   the network into recovery domains based on physical network topology,
   control plane capabilities, or various traffic engineering
   constraints.

   An example often addressed in the literature is the metro-core-metro
   application (sometimes extended to a metro-metro/core-core) within a
   single transport layer (see Section 7.2).  For such a case, an end-
   to-end LSP is defined between the ingress and egress metro nodes,
   while LSP segments may be defined within the metro or core sub-
   networks.  Each of these topological structures determines a so-
   called "recovery domain" since each of the LSPs they carry can have
   its own recovery type (or even scheme).  The support of multiple
   recovery types and schemes within a sub-network is referred to as a
   "multi-recovery capable domain" or simply "multi-recovery domain".

7.2.  Vertical Hierarchy (Layers)

   It is very challenging to combine the different recovery capabilities
   available across the path (i.e., switching capable) and section
   layers to ensure that certain network survivability objectives are
   met for the network-supported services.

Papadimitriou & Mannie       Informational                     [Page 28]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   As a first analysis step, one can draw the following guidelines for
   a vertical coordination of the recovery mechanisms:

   - The lower the layer, the faster the notification and switching.

   - The higher the layer, the finer the granularity of the recoverable
     entity and therefore the granularity of the recovery resource.

   Moreover, in the context of this analysis, a vertical hierarchy
   consists of multiple layered transport planes providing different:

   - Discrete bandwidth granularities for non-packet LSPs such as OCh,
     ODUk, STS_SPE/HOVC, and VT_SPE/LOVC LSPs and continuous bandwidth
     granularities for packet LSPs.

   - Potential recovery capabilities with different temporal
     granularities: ranging from milliseconds to tens of seconds

   Note: based on the bandwidth granularity, we can determine four
   classes of vertical hierarchies: (1) packet over packet, (2) packet
   over circuit, (3) circuit over packet, and (4) circuit over circuit.
   Below we briefly expand on (4) only. (2) is covered in [RFC3386]. (1)
   is extensively covered by the MPLS Working Group, and (3) by the PWE3
   Working Group.

   In SONET/SDH environments, one typically considers the VT_SPE/LOVC
   and STS SPE/HOVC as independent layers (for example, VT_SPE/LOVC LSP
   uses the underlying STS_SPE/HOVC LSPs as links).  In OTN, the ODUk
   path layers will lie on the OCh path layer, i.e., the ODUk LSPs use
   the underlying OCh LSPs as OTUk links.  Note here that lower layer
   LSPs may simply be provisioned and not necessarily dynamically
   triggered or established (control driven approach).  In this context,
   an LSP at the path layer (i.e., established using GMPLS signaling),
   such as an optical channel LSP, appears at the OTUk layer as a link,
   controlled by a link management protocol such as LMP.

   The first key issue with multi-layer recovery is that achieving
   individual or bulk LSP recovery will be as efficient as the
   underlying link (local span) recovery.  In such a case, the span can
   be either protected or unprotected, but the LSP it carries must be
   (at least locally) recoverable.  Therefore, the span recovery process
   can be either independent when protected (or restorable), or
   triggered by the upper LSP recovery process.  The former case
   requires coordination to achieve subsequent LSP recovery.  Therefore,
   in order to achieve robustness and fast convergence, multi-layer
   recovery requires a fine-tuned coordination mechanism.

Papadimitriou & Mannie       Informational                     [Page 29]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   Moreover, in the absence of adequate recovery mechanism coordination
   (for instance, a pre-determined coordination when using a hold-off
   timer), a failure notification may propagate from one layer to the
   next one within a recovery hierarchy.  This can cause "collisions"
   and trigger simultaneous recovery actions that may lead to race
   conditions and, in turn, reduce the optimization of the resource
   utilization and/or generate global instabilities in the network (see
   [MANCHESTER]).  Therefore, a consistent and efficient escalation
   strategy is needed to coordinate recovery across several layers.

   One can expect that the definition of the recovery mechanisms and
   protocol(s) is technology-independent so that they can be
   consistently implemented at different layers; this would in turn
   simplify their global coordination.  Moreover, as mentioned in
   [RFC3386], some looser form of coordination and communication between
   (vertical) layers such as a consistent hold-off timer configuration
   (and setup through signaling during the working LSP establishment)
   can be considered, thereby allowing the synchronization between
   recovery actions performed across these layers.

7.2.1.  Recovery Granularity

   In most environments, the design of the network and the vertical
   distribution of the LSP bandwidth are such that the recovery
   granularity is finer at higher layers.  The OTN and SONET/SDH layers
   can recover only the whole section or the individual connections they
   transports whereas the IP/MPLS control plane can recover individual
   packet LSPs or groups of packet LSPs independently of their
   granularity.  On the other side, the recovery granularity at the
   sub-wavelength level (i.e., SONET/SDH) can be provided only when the
   network includes devices switching at the same granularity (and thus
   not with optical channel level).  Therefore, the network layer can
   deliver control-plane-driven recovery mechanisms on a per-LSP basis
   if and only if these LSPs have their corresponding switching
   granularity supported at the transport plane level.

7.3.  Escalation Strategies

   There are two types of escalation strategies (see [DEMEESTER]):
   bottom-up and top-down.

   The bottom-up approach assumes that lower layer recovery types and
   schemes are more expedient and faster than upper layer ones.
   Therefore, we can inhibit or hold off higher layer recovery.
   However, this assumption is not entirely true.  Consider for instance
   a SONET/SDH based protection mechanism (with a protection switching
   time of less than 50 ms) lying on top of an OTN restoration mechanism
   (with a restoration time of less than 200 ms).  Therefore, this

Papadimitriou & Mannie       Informational                     [Page 30]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   assumption should be (at least) clarified as: the lower layer
   recovery mechanism is expected to be faster than the upper level one,
   if the same type of recovery mechanism is used at each layer.

   Consequently, taking into account the recovery actions at the
   different layers in a bottom-up approach: if lower layer recovery
   mechanisms are provided and sequentially activated in conjunction
   with higher layer ones, the lower layers must have an opportunity to
   recover normal traffic before the higher layers do.  However, if
   lower layer recovery is slower than higher layer recovery, the lower
   layer must either communicate the failure-related information to the
   higher layer(s) (and allow it to perform recovery), or use a hold-off
   timer in order to temporarily set the higher layer recovery action in
   a "standby mode".  Note that the a priori information exchange
   between layers concerning their efficiency is not within the current
   scope of this document.  Nevertheless, the coordination functionality
   between layers must be configurable and tunable.

   For example, coordination between the optical and packet layer
   control plane enables the optical layer to perform the failure
   management operations (in particular, failure detection and
   notification) while giving to the packet layer control plane the
   authority to decide and perform the recovery actions.  If the packet
   layer recovery action is unsuccessful, fallback at the optical layer
   can be performed subsequently.

   The top-down approach attempts service recovery at the higher layers
   before invoking lower layer recovery.  Higher layer recovery is
   service selective, and permits "per-CoS" or "per-connection" re-
   routing.  With this approach, the most important aspect is that the
   upper layer should provide its own reliable and independent failure
   detection mechanism from the lower layer.

   [DEMEESTER] also suggests recovery mechanisms incorporating a
   coordinated effort shared by two adjacent layers with periodic status
   updates.  Moreover, some of these recovery operations can be pre-
   assigned (on a per-link basis) to a certain layer, e.g., a given link
   will be recovered at the packet layer while another will be recovered
   at the optical layer.

7.4.  Disjointness

   Having link and node diverse working and recovery LSPs/spans does not
   guarantee their complete disjointness.  Due to the common physical
   layer topology (passive), additional hierarchical concepts, such as
   the Shared Risk Link Group (SRLG), and mechanisms, such as SRLG
   diverse path computation, must be developed to provide complete
   working and recovery LSP/span disjointness (see [IPO-IMP] and

Papadimitriou & Mannie       Informational                     [Page 31]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   [RFC4202]).  Otherwise, a failure affecting the working LSP/span
   would also potentially affect the recovery LSP/span; one refers to
   such an event as "common failure".

7.4.1.  SRLG Disjointness

   A Shared Risk Link Group (SRLG) is defined as the set of links
   sharing a common risk (such as a common physical resource such as a
   fiber link or a fiber cable).  For instance, a set of links L belongs
   to the same SRLG s, if they are provisioned over the same fiber link
   f.

   The SRLG properties can be summarized as follows:

   1) A link belongs to more than one SRLG if and only if it crosses one
      of the resources covered by each of them.

   2) Two links belonging to the same SRLG can belong individually to
      (one or more) other SRLGs.

   3) The SRLG set S of an LSP is defined as the union of the individual
      SRLG s of the individual links composing this LSP.

   SRLG disjointness is also applicable to LSPs:

      The LSP SRLG disjointness concept is based on the following
      postulate: an LSP (i.e., a sequence of links and nodes) covers an
      SRLG if and only if it crosses one of the links or nodes belonging
      to that SRLG.

      Therefore, the SRLG disjointness for LSPs, can be defined as
      follows: two LSPs are disjoint with respect to an SRLG s if and
      only if they do not cover simultaneously this SRLG s.

      Whilst the SRLG disjointness for LSPs with respect to a set S of
      SRLGs, is defined as follows: two LSPs are disjoint with respect
      to a set of SRLGs S if and only if the set of SRLGs that are
      common to both LSPs is disjoint from set S.

   The impact on recovery is noticeable: SRLG disjointness is a
   necessary (but not a sufficient) condition to ensure network
   survivability.  With respect to the physical network resources, a
   working-recovery LSP/span pair must be SRLG-disjoint in case of
   dedicated recovery type.  On the other hand, in case of shared
   recovery, a group of working LSP/spans must be mutually SRLG-disjoint
   in order to allow for a (single and common) shared recovery LSP that
   is itself SRLG-disjoint from each of the working LSPs/spans.

Papadimitriou & Mannie       Informational                     [Page 32]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

8.  Recovery Mechanisms Analysis

   In order to provide a structured analysis of the recovery mechanisms
   detailed in the previous sections, the following dimensions can be
   considered:

   1. Fast convergence (performance): provide a mechanism that
      aggregates multiple failures (implying fast failure detection and
      correlation mechanisms) and fast recovery decision independently
      of the number of failures occurring in the optical network (also
      implying a fast failure notification).

   2. Efficiency (scalability): minimize the switching time required for
      LSP/span recovery independently of the number of LSPs/spans being
      recovered (this implies efficient failure correlation, fast
      failure notification, and time-efficient recovery mechanisms).

   3. Robustness (availability): minimize the LSP/span downtime
      independently of the underlying topology of the transport plane
      (this implies a highly responsive recovery mechanism).

   4. Resource optimization (optimality): minimize the resource
      capacity, including LSPs/spans and nodes (switching capacity),
      required for recovery purposes; this dimension can also be
      referred to as optimizing the sharing degree of the recovery
      resources.

   5. Cost optimization: provide a cost-effective recovery type/scheme.

   However, these dimensions are either outside the scope of this
   document (such as cost optimization and recovery path computational
   aspects) or mutually conflicting.  For instance, it is obvious that
   providing a 1+1 LSP protection minimizes the LSP downtime (in case of
   failure) while being non-scalable and consuming recovery resource
   without enabling any extra-traffic.

   The following sections analyze the recovery phases and mechanisms
   detailed in the previous sections with respect to the dimensions
   described above in order to assess the GMPLS protocol suite
   capabilities and applicability.  In turn, this allows the evaluation
   of the potential need for further GMPLS signaling and routing
   extensions.

Papadimitriou & Mannie       Informational                     [Page 33]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

8.1.  Fast Convergence (Detection/Correlation and Hold-off Time)

   Fast convergence is related to the failure management operations.  It
   refers to the time elapsed between failure detection/correlation and
   hold-off time, the point at which the recovery switching actions are
   initiated.  This point has been detailed in Section 4.

8.2.  Efficiency (Recovery Switching Time)

   In general, the more pre-assignment/pre-planning of the recovery
   LSP/span, the more rapid the recovery is.  Because protection implies
   pre-assignment (and cross-connection) of the protection resources, in
   general, protection recovers faster than restoration.

   Span restoration is likely to be slower than most span protection
   types; however this greatly depends on the efficiency of the span
   restoration signaling.  LSP restoration with pre-signaled and pre-
   selected recovery resources is likely to be faster than fully dynamic
   LSP restoration, especially because of the elimination of any
   potential crankback during the recovery LSP establishment.

   If one excludes the crankback issue, the difference between dynamic
   and pre-planned restoration depends on the restoration path
   computation and selection time.  Since computational considerations
   are outside the scope of this document, it is up to the vendor to
   determine the average and maximum path computation time in different
   scenarios and to the operator to decide whether or not dynamic
   restoration is advantageous over pre-planned schemes that depend on
   the network environment.  This difference also depends on the
   flexibility provided by pre-planned restoration versus dynamic
   restoration.  Pre-planned restoration implies a somewhat limited
   number of failure scenarios (that can be due, for instance, to local
   storage capacity limitation).  Dynamic restoration enables on-demand
   path computation based on the information received through failure
   notification message, and as such, it is more robust with respect to
   the failure scenario scope.

   Moreover, LSP segment restoration, in particular, dynamic restoration
   (i.e., no path pre-computation, so none of the recovery resource is
   pre-reserved) will generally be faster than end-to-end LSP
   restoration.  However, local LSP restoration assumes that each LSP
   segment end-point has enough computational capacity to perform this
   operation while end-to-end LSP restoration requires only that LSP
   end-points provide this path computation capability.

   Recovery time objectives for SONET/SDH protection switching (not
   including time to detect failure) are specified in [G.841] at 50 ms,
   taking into account constraints on distance, number of connections

Papadimitriou & Mannie       Informational                     [Page 34]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   involved, and in the case of ring enhanced protection, number of
   nodes in the ring.  Recovery time objectives for restoration
   mechanisms have been proposed through a separate effort [RFC3386].

8.3.  Robustness

   In general, the less pre-assignment (protection)/pre-planning
   (restoration) of the recovery LSP/span, the more robust the recovery
   type or scheme is to a variety of single failures, provided that
   adequate resources are available.  Moreover, the pre-selection of the
   recovery resources gives (in the case of multiple failure scenarios)
   less flexibility than no recovery resource pre-selection.  For
   instance, if failures occur that affect two LSPs sharing a common
   link along their restoration paths, then only one of these LSPs can
   be recovered.  This occurs unless the restoration path of at least
   one of these LSPs is re-computed, or the local resource assignment is
   modified on the fly.

   In addition, recovery types and schemes with pre-planned recovery
   resources (in particular, LSP/spans for protection and LSPs for
   restoration purposes) will not be able to recover from failures that
   simultaneously affect both the working and recovery LSP/span.  Thus,
   the recovery resources should ideally be as disjoint as possible
   (with respect to link, node, and SRLG) from the working ones, so that
   any single failure event will not affect both working and recovery
   LSP/span.  In brief, working and recovery resources must be fully
   diverse in order to guarantee that a given failure will not affect
   simultaneously the working and the recovery LSP/span.  Also, the risk
   of simultaneous failure of the working and the recovery LSPs can be
   reduced.  It is reduced by computing a new recovery path whenever a
   failure occurs along one of the recovery LSPs or by computing a new
   recovery path and provision the corresponding LSP whenever a failure
   occurs along a working LSP/span.  Both methods enable the network to
   maintain the number of available recovery path constant.

   The robustness of a recovery scheme is also determined by the amount
   of pre-reserved (i.e., signaled) recovery resources within a given
   shared resource pool: as the sharing degree of recovery resources
   increases, the recovery scheme becomes less robust to multiple
   LSP/span failure occurrences.  Recovery schemes, in particular
   restoration, with pre-signaled resource reservation (with or without
   pre-selection) should be capable of reserving an adequate amount of
   resource to ensure recovery from any specific set of failure events,
   such as any single SRLG failure, any two SRLG failures, etc.

Papadimitriou & Mannie       Informational                     [Page 35]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

8.4.  Resource Optimization

   It is commonly admitted that sharing recovery resources provides
   network resource optimization.  Therefore, from a resource
   utilization perspective, protection schemes are often classified with
   respect to their degree of sharing recovery resources with the
   working entities.  Moreover, non-permanent bridging protection types
   allow (under normal conditions) for extra-traffic over the recovery
   resources.

   From this perspective, the following statements are true:

   1) 1+1 LSP/Span protection is the most resource-consuming protection
      type because it does not allow for any extra traffic.

   2) 1:1 LSP/span recovery requires dedicated recovery LSP/span
      allowing for extra traffic.

   3) 1:N and M:N LSP/span recovery require 1 (and M, respectively)
      recovery LSP/span (shared between the N working LSP/span) allowing
      for extra traffic.

   Obviously, 1+1 protection precludes, and 1:1 recovery does not allow
   for any recovery LSP/span sharing, whereas 1:N and M:N recovery do
   allow sharing of 1 (M, respectively) recovery LSP/spans between N
   working LSP/spans.  However, despite the fact that 1:1 LSP recovery
   precludes the sharing of the recovery LSP, the recovery schemes that
   can be built from it (e.g., (1:1)^n, see Section 5.4) do allow
   sharing of its recovery resources.  In addition, the flexibility in
   the usage of shared recovery resources (in particular, shared links)
   may be limited because of network topology restrictions, e.g., fixed
   ring topology for traditional enhanced protection schemes.

   On the other hand, when using LSP restoration with pre-signaled
   resource reservation, the amount of reserved restoration capacity is
   determined by the local bandwidth reservation policies.  In LSP
   restoration schemes with re-provisioning, a pool of spare resources
   can be defined from which all resources are selected after failure
   occurrence for the purpose of restoration path computation.  The
   degree to which restoration schemes allow sharing amongst multiple
   independent failures is then directly inferred from the size of the
   resource pool.  Moreover, in all restoration schemes, spare resources
   can be used to carry preemptible traffic (thus over preemptible
   LSP/span) when the corresponding resources have not been committed
   for LSP/span recovery purposes.

   From this, it clearly follows that less recovery resources (i.e.,
   LSP/spans and switching capacity) have to be allocated to a shared

Papadimitriou & Mannie       Informational                     [Page 36]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   recovery resource pool if a greater sharing degree is allowed.  Thus,
   the network survivability level is determined by the policy that
   defines the amount of shared recovery resources and by the maximum
   sharing degree allowed for these recovery resources.

8.4.1.  Recovery Resource Sharing

   When recovery resources are shared over several LSP/Spans, the use of
   the Maximum Reservable Bandwidth, the Unreserved Bandwidth, and the
   Maximum LSP Bandwidth (see [RFC4202]) provides the information needed
   to obtain the optimization of the network resources allocated for
   shared recovery purposes.

   The Maximum Reservable Bandwidth is defined as the Maximum Link
   Bandwidth but it may be greater in case of link over-subscription.

   The Unreserved Bandwidth (at priority p) is defined as the bandwidth
   not yet reserved on a given TE link (its initial value for each
   priority p corresponds to the Maximum Reservable Bandwidth).  Last,
   the Maximum LSP Bandwidth (at priority p) is defined as the smaller
   of Unreserved Bandwidth (at priority p) and Maximum Link Bandwidth.

   Here, one generally considers a recovery resource sharing degree (or
   ratio) to globally optimize the shared recovery resource usage.  The
   distribution of the bandwidth utilization per TE link can be inferred
   from the per-priority bandwidth pre-allocation.  By using the Maximum
   LSP Bandwidth and the Maximum Reservable Bandwidth, the amount of
   (over-provisioned) resources that can be used for shared recovery
   purposes is known from the IGP.

   In order to analyze this behavior, we define the difference between
   the Maximum Reservable Bandwidth (in the present case, this value is
   greater than the Maximum Link Bandwidth) and the Maximum LSP
   Bandwidth per TE link i as the Maximum Shareable Bandwidth or
   max_R[i].  Within this quantity, the amount of bandwidth currently
   allocated for shared recovery per TE link i is defined as R[i].  Both
   quantities are expressed in terms of discrete bandwidth units (and
   thus, the Minimum LSP Bandwidth is of one bandwidth unit).

   The knowledge of this information available per TE link can be
   exploited in order to optimize the usage of the resources allocated
   per TE link for shared recovery.  If one refers to r[i] as the actual
   bandwidth per TE link i (in terms of discrete bandwidth units)
   committed for shared recovery, then the following quantity must be
   maximized over the potential TE link candidates:

        sum {i=1}^N [(R{i} - r{i})/(t{i} - b{i})]

Papadimitriou & Mannie       Informational                     [Page 37]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

        or equivalently: sum {i=1}^N [(R{i} - r{i})/r{i}]

        with R{i} >= 1 and r{i} >= 1 (in terms of per component
        bandwidth unit)

   In this formula, N is the total number of links traversed by a given
   LSP, t[i] the Maximum Link Bandwidth per TE link i, and b[i] the sum
   per TE link i of the bandwidth committed for working LSPs and other
   recovery LSPs (thus except "shared bandwidth" LSPs).  The quantity
   [(R{i} - r{i})/r{i}] is defined as the Shared (Recovery) Bandwidth
   Ratio per TE link i.  In addition, TE links for which R[i] reaches
   max_R[i] or for which r[i] = 0 are pruned during shared recovery path
   computation as well as TE links for which max_R[i] = r[i] that can
   simply not be shared.

   More generally, one can draw the following mapping between the
   available bandwidth at the transport and control plane level:

                                 - ---------- Max Reservable Bandwidth
                                |  -----  ^
                                |R -----  |
                                |  -----  |
                                 - -----  |max_R
                                   -----  |
   --------  TE link Capacity    - ------ | - Maximum TE Link Bandwidth
   -----                        |r -----  v
   -----     <------ b ------>   - ---------- Maximum LSP Bandwidth
   -----                           -----
   -----                           -----
   -----                           -----
   -----                           -----
   -----                           ----- <--- Minimum LSP Bandwidth
   -------- 0                      ---------- 0

   Note that the above approach does not require the flooding of any per
   LSP information or any detailed distribution of the bandwidth
   allocation per component link or individual ports or even any per-
   priority shareable recovery bandwidth information (using a dedicated
   sub-TLV).  The latter would provide the same capability as the
   already defined Maximum LSP bandwidth per-priority information.  This
   approach is referred to as a Partial (or Aggregated) Information
   Routing as described in [KODIALAM1] and [KODIALAM2].  They show that
   the difference obtained with a Full (or Complete) Information Routing
   approach (where for the whole set of working and recovery LSPs, the
   amount of bandwidth units they use per-link is known at each node and
   for each link) is clearly negligible.  The Full Information Routing

Papadimitriou & Mannie       Informational                     [Page 38]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   approach is detailed in [GLI].  Note also that both approaches rely
   on the deterministic knowledge (at different degrees) of the network
   topology and resource usage status.

   Moreover, extending the GMPLS signaling capabilities can enhance the
   Partial Information Routing approach.  It is enhanced by allowing
   working-LSP-related information and, in particular, its path
   (including link and node identifiers) to be exchanged with the
   recovery LSP request.  This enables more efficient admission control
   at upstream nodes of shared recovery resources, and in particular,
   links (see Section 8.4.3).

8.4.2.  Recovery Resource Sharing and SRLG Recovery

   Resource shareability can also be maximized with respect to the
   number of times each SRLG is protected by a recovery resource (in
   particular, a shared TE link) and methods can be considered for
   avoiding contention of the shared recovery resources in case of
   single SRLG failure.  These methods enable the sharing of recovery
   resources between two (or more) recovery LSPs, if their respective
   working LSPs are mutually disjoint with respect to link, node, and
   SRLGs.  Then, a single failure does not simultaneously disrupt
   several (or at least two) working LSPs.

   For instance, [BOUILLET] shows that the Partial Information Routing
   approach can be extended to cover recovery resource shareability with
   respect to SRLG recoverability (i.e., the number of times each SRLG
   is recoverable).  By flooding this aggregated information per TE
   link, path computation and selection of SRLG-diverse recovery LSPs
   can be optimized with respect to the sharing of recovery resource
   reserved on each TE link.  This yields a performance difference of
   less than 5%, which is negligible compared to the corresponding Full
   Information Flooding approach (see [GLI]).

   For this purpose, additional extensions to [RFC4202] in support of
   path computation for shared mesh recovery have been often considered
   in the literature.  TE link attributes would include, among others,
   the current number of recovery LSPs sharing the recovery resources
   reserved on the TE link, and the current number of SRLGs recoverable
   by this amount of (shared) recovery resources reserved on the TE
   link.  The latter is equivalent to the current number of SRLGs that
   will be recovered by the recovery LSPs sharing the recovery resource
   reserved on the TE link.  Then, if explicit SRLG recoverability is
   considered, a TE link attribute would be added that includes the
   explicit list of SRLGs (recoverable by the shared recovery resource
   reserved on the TE link) and their respective shareable recovery
   bandwidths.  The latter information is equivalent to the shareable
   recovery bandwidth per SRLG (or per group of SRLGs), which implies

Papadimitriou & Mannie       Informational                     [Page 39]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   that the amount of shareable bandwidth and the number of listed SRLGs
   will decrease over time.

   Compared to the case of recovery resource sharing only (regardless of
   SRLG recoverability, as described in Section 8.4.1), these additional
   TE link attributes would potentially deliver better path computation
   and selection (at a distinct ingress node) for shared mesh recovery
   purposes.  However, due to the lack of evidence of better efficiency
   and due to the complexity that such extensions would generate, they
   are not further considered in the scope of the present analysis.  For
   instance, a per-SRLG group minimum/maximum shareable recovery
   bandwidth is restricted by the length that the corresponding (sub-)
   TLV may take and thus the number of SRLGs that it can include.
   Therefore, the corresponding parameter should not be translated into
   GMPLS routing (or even signaling) protocol extensions in the form of
   TE link sub-TLV.

8.4.3.  Recovery Resource Sharing, SRLG Disjointness and Admission
        Control

   Admission control is a strict requirement to be fulfilled by nodes
   giving access to shared links.  This can be illustrated using the
   following network topology:

      A ------ C ====== D
      |        |        |
      |        |        |
      |        B        |
      |        |        |
      |        |        |
       ------- E ------ F

   Node A creates a working LSP to D (A-C-D), B creates simultaneously a
   working LSP to D (B-C-D) and a recovery LSP (B-E-F-D) to the same
   destination.  Then, A decides to create a recovery LSP to D (A-E-F-
   D), but since the C-D span carries both working LSPs, node E should
   either assign a dedicated resource for this recovery LSP or reject
   this request if the C-D span has already reached its maximum recovery
   bandwidth sharing ratio.  In the latter case, C-D span failure would
   imply that one of the working LSP would not be recoverable.

   Consequently, node E must have the required information to perform
   admission control for the recovery LSP requests it processes
   (implying for instance, that the path followed by the working LSP is
   carried with the corresponding recovery LSP request).  If node E can
   guarantee that the working LSPs (A-C-D and B-C-D) are SRLG disjoint
   over the C-D span, it may securely accept the incoming recovery LSP
   request and assign to the recovery LSPs (A-E-F-D and B-E-F-D) the

Papadimitriou & Mannie       Informational                     [Page 40]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   same resources on the link E-F.  This may occur if the link E-F has
   not yet reached its maximum recovery bandwidth sharing ratio.  In
   this example, one assumes that the node failure probability is
   negligible compared to the link failure probability.

   To achieve this, the path followed by the working LSP is transported
   with the recovery LSP request and examined at each upstream node of
   potentially shareable links.  Admission control is performed using
   the interface identifiers (included in the path) to retrieve in the
   TE DataBase the list of SRLG IDs associated to each of the working
   LSP links.  If the working LSPs (A-C-D and B-C-D) have one or more
   link or SRLG ID in common (in this example, one or more SRLG id in
   common over the span C-D), node E should not assign the same resource
   over link E-F to the recovery LSPs (A-E-F-D and B-E-F-D).  Otherwise,
   one of these working LSPs would not be recoverable if C-D span
   failure occurred.

   There are some issues related to this method; the major one is the
   number of SRLG IDs that a single link can cover (more than 100, in
   complex environments).  Moreover, when using link bundles, this
   approach may generate the rejection of some recovery LSP requests.
   This occurs when the SRLG sub-TLV corresponding to a link bundle
   includes the union of the SRLG id list of all the component links
   belonging to this bundle (see [RFC4202] and [RFC4201]).

   In order to overcome this specific issue, an additional mechanism may
   consist of querying the nodes where the information would be
   available (in this case, node E would query C).  The main drawback of
   this method is that (in addition to the dedicated mechanism(s) it
   requires) it may become complex when several common nodes are
   traversed by the working LSPs.  Therefore, when using link bundles,
   solving this issue is closely related to the sequence of the recovery
   operations.  Per-component flooding of SRLG identifiers would deeply
   impact the scalability of the link state routing protocol.
   Therefore, one may rely on the usage of an on-line accessible network
   management system.

Papadimitriou & Mannie       Informational                     [Page 41]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

9.  Summary and Conclusions

   The following table summarizes the different recovery types and
   schemes analyzed throughout this document.

   --------------------------------------------------------------------
              |       Path Search (computation and selection)
   --------------------------------------------------------------------
              |       Pre-planned (a)      |         Dynamic (b)
   --------------------------------------------------------------------
          |   | faster recovery            | Does not apply
          |   | less flexible              |
          | 1 | less robust                |
          |   | most resource-consuming    |
   Path   |   |                            |
   Setup   ------------------------------------------------------------
          |   | relatively fast recovery   | Does not apply
          |   | relatively flexible        |
          | 2 | relatively robust          |
          |   | resource consumption       |
          |   |  depends on sharing degree |
           ------------------------------------------------------------
          |   | relatively fast recovery   | less faster (computation)
          |   | more flexible              | most flexible
          | 3 | relatively robust          | most robust
          |   | less resource-consuming    | least resource-consuming
          |   |  depends on sharing degree |
   --------------------------------------------------------------------

   1a. Recovery LSP setup (before failure occurrence) with resource
       reservation (i.e., signaling) and selection is referred to as LSP
       protection.

   2a. Recovery LSP setup (before failure occurrence) with resource
       reservation (i.e., signaling) and with resource pre-selection is
       referred to as pre-planned LSP re-routing with resource pre-
       selection.  This implies only recovery LSP activation after
       failure occurrence.

   3a. Recovery LSP setup (before failure occurrence) with resource
       reservation (i.e., signaling) and without resource selection is
       referred to as pre-planned LSP re-routing without resource pre-
       selection.  This implies recovery LSP activation and resource
       (i.e., label) selection after failure occurrence.

   3b. Recovery LSP setup after failure occurrence is referred to as to
       as LSP re-routing, which is full when recovery LSP path
       computation occurs after failure occurrence.

Papadimitriou & Mannie       Informational                     [Page 42]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   Thus, the term pre-planned refers to recovery LSP path pre-
   computation, signaling (reservation), and a priori resource selection
   (optional), but not cross-connection.  Also, the shared-mesh recovery
   scheme can be viewed as a particular case of 2a) and 3a), using the
   additional constraint described in Section 8.4.3.

   The implementation of these recovery mechanisms requires only
   considering extensions to GMPLS signaling protocols (i.e., [RFC3471]
   and [RFC3473]).  These GMPLS signaling extensions should mainly focus
   in delivering (1) recovery LSP pre-provisioning for the cases 1a, 2a,
   and 3a, (2) LSP failure notification, (3) recovery LSP switching
   action(s), and (4) reversion mechanisms.

   Moreover, the present analysis (see Section 8) shows that no GMPLS
   routing extensions are expected to efficiently implement any of these
   recovery types and schemes.

10.  Security Considerations

   This document does not introduce any additional security issue or
   imply any specific security consideration from [RFC3945] to the
   current RSVP-TE GMPLS signaling, routing protocols (OSPF-TE, IS-IS-
   TE) or network management protocols.

   However, the authorization of requests for resources by GMPLS-capable
   nodes should determine whether a given party, presumably already
   authenticated, has a right to access the requested resources.  This
   determination is typically a matter of local policy control, for
   example, by setting limits on the total bandwidth made available to
   some party in the presence of resource contention.  Such policies may
   become quite complex as the number of users, types of resources, and
   sophistication of authorization rules increases.  This is
   particularly the case for recovery schemes that assume pre-planned
   sharing of recovery resources, or contention for resources in case of
   dynamic re-routing.

   Therefore, control elements should match the requests against the
   local authorization policy.  These control elements must be capable
   of making decisions based on the identity of the requester, as
   verified cryptographically and/or topologically.

11.  Acknowledgements

   The authors would like to thank Fabrice Poppe (Alcatel) and Bart
   Rousseau (Alcatel) for their revision effort, and Richard Rabbat
   (Fujitsu Labs), David Griffith (NIST), and Lyndon Ong (Ciena) for
   their useful comments.

Papadimitriou & Mannie       Informational                     [Page 43]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   Thanks also to Adrian Farrel for the thorough review of the document.

12.  References

12.1.  Normative References

   [RFC2119]    Bradner, S., "Key words for use in RFCs to Indicate
                Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3471]    Berger, L., "Generalized Multi-Protocol Label Switching
                (GMPLS) Signaling Functional Description", RFC 3471,
                January 2003.

   [RFC3473]    Berger, L., "Generalized Multi-Protocol Label Switching
                (GMPLS) Signaling Resource ReserVation Protocol-Traffic
                Engineering (RSVP-TE) Extensions", RFC 3473, January
                2003.

   [RFC3945]    Mannie, E., "Generalized Multi-Protocol Label Switching
                (GMPLS) Architecture", RFC 3945, October 2004.

   [RFC4201]    Kompella, K., Rekhter, Y., and L. Berger, "Link Bundling
                in MPLS Traffic Engineering (TE)", RFC 4201, October
                2005.

   [RFC4202]    Kompella, K., Ed. and Y. Rekhter, Ed., "Routing
                Extensions in Support of Generalized Multi-Protocol
                Label Switching (GMPLS)", RFC 4202, October 2005.

   [RFC4204]    Lang, J., Ed., "Link Management Protocol (LMP)", RFC
                4204, October 2005.

   [RFC4209]    Fredette, A., Ed. and J. Lang, Ed., "Link Management
                Protocol (LMP) for Dense Wavelength Division
                Multiplexing (DWDM) Optical Line Systems", RFC 4209,
                October 2005.

   [RFC4427]    Mannie E., Ed. and D. Papadimitriou, Ed., "Recovery
                (Protection and Restoration) Terminology for Generalized
                Multi-Protocol Label Switching (GMPLS)", RFC 4427, March
                2006.

12.2.  Informative References

   [BOUILLET]   E. Bouillet, et al., "Stochastic Approaches to Compute
                Shared Meshed Restored Lightpaths in Optical Network
                Architectures," IEEE Infocom 2002, New York City, June
                2002.

Papadimitriou & Mannie       Informational                     [Page 44]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   [DEMEESTER]  P. Demeester, et al., "Resilience in Multilayer
                Networks," IEEE Communications Magazine, Vol. 37, No. 8,
                pp. 70-76, August 1998.

   [GLI]        G. Li, et al., "Efficient Distributed Path Selection for
                Shared Restoration Connections," IEEE Infocom 2002, New
                York City, June 2002.

   [IPO-IMP]    Strand, J. and A. Chiu, "Impairments and Other
                Constraints on Optical Layer Routing", RFC 4054, May
                2005.

   [KODIALAM1]  M. Kodialam and T.V. Lakshman, "Restorable Dynamic
                Quality of Service Routing," IEEE Communications
                Magazine, pp. 72-81, June 2002.

   [KODIALAM2]  M. Kodialam and T.V. Lakshman, "Dynamic Routing of
                Restorable Bandwidth-Guaranteed Tunnels using Aggregated
                Network Resource Usage Information," IEEE/ ACM
                Transactions on Networking, pp. 399-410, June 2003.

   [MANCHESTER] J. Manchester, P. Bonenfant and C. Newton, "The
                Evolution of Transport Network Survivability," IEEE
                Communications Magazine, August 1999.

   [RFC3386]    Lai, W. and D. McDysan, "Network Hierarchy and
                Multilayer Survivability", RFC 3386, November 2002.

   [T1.105]     ANSI, "Synchronous Optical Network (SONET): Basic
                Description Including Multiplex Structure, Rates, and
                Formats," ANSI T1.105, January 2001.

   [WANG]       J. Wang, L. Sahasrabuddhe, and B. Mukherjee, "Path vs.
                Subpath vs. Link Restoration for Fault Management in
                IP-over-WDM Networks: Performance Comparisons Using
                GMPLS Control Signaling," IEEE Communications Magazine,
                pp. 80-87, November 2002.

   For information on the availability of the following documents,
   please see http://www.itu.int

   [G.707]      ITU-T, "Network Node Interface for the Synchronous
                Digital Hierarchy (SDH)," Recommendation G.707, October
                2000.

   [G.709]      ITU-T, "Network Node Interface for the Optical Transport
                Network (OTN)," Recommendation G.709, February 2001 (and
                Amendment no.1, October 2001).

Papadimitriou & Mannie       Informational                     [Page 45]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

   [G.783]      ITU-T, "Characteristics of Synchronous Digital Hierarchy
                (SDH) Equipment Functional Blocks," Recommendation
                G.783, October 2000.

   [G.798]      ITU-T, "Characteristics of optical transport network
                hierarchy equipment functional block," Recommendation
                G.798, June 2004.

   [G.806]      ITU-T, "Characteristics of Transport Equipment -
                Description Methodology and Generic Functionality",
                Recommendation G.806, October 2000.

   [G.841]      ITU-T, "Types and Characteristics of SDH Network
                Protection Architectures," Recommendation G.841, October
                1998.

   [G.842]      ITU-T, "Interworking of SDH network protection
                architectures," Recommendation G.842, October 1998.

   [G.874]      ITU-T, "Management aspects of the optical transport
                network element," Recommendation G.874, November 2001.

Editors' Addresses

   Dimitri Papadimitriou
   Alcatel
   Francis Wellesplein, 1
   B-2018 Antwerpen, Belgium

   Phone:  +32 3 240-8491
   EMail: dimitri.papadimitriou@alcatel.be

   Eric Mannie
   Perceval
   Rue Tenbosch, 9
   1000 Brussels
   Belgium

   Phone: +32-2-6409194
   EMail: eric.mannie@perceval.net

Papadimitriou & Mannie       Informational                     [Page 46]



RFC 4428               GMPLS Recovery Mechanisms              March 2006

Full Copyright Statement

   Copyright (C) The Internet Society (2006).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
   ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
   INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
   INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
   WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Intellectual Property

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at
   ietf-ipr@ietf.org.

Acknowledgement

   Funding for the RFC Editor function is provided by the IETF
   Administrative Support Activity (IASA).

Papadimitriou & Mannie       Informational                     [Page 47]